Limiting GPT knowledge to specific web-sources / web-links

Hi Everyone!
I am building a GPT that provides answers to questions based only on the information provided in a specific webpage. Still the GPT uses “bing” and does not really come up with the information provided on the page, sometimes it gives information that is not really available from the specific sourcs.
Does anyone have an idea how to solve the problem ie. how to limit the source information to a specific website?
thannks
felix

Current Promt:
you are an employee of “Investment Consulting” and give financial forecasts and make outlook predictions based solely on the following web link and all sub-pages:
source: specific_website_url

You are not allowed to use bing or to use any other ressource as information. Also do not use any other data - only answer based on your specific knowledge from the above given link. If nothing is mentioned in the link, say that you do not have a foreceast currently.

Bing is the only way that GPTs have of accessing the internet, be it a URL or a search. If you want direct fetching you would need to use the Assistants API, or add a function in the GPT that fetches information (you’ll need a middleware or use a different 3rd party service).

You can later refine the prompt in the instructions to limit its behavior. I recommend also adding handling of edge cases like the site being dynamic (loaded with JS and thus not directly fetchable) and the page being empty, behind auth, or in different languages.